Python 3.6 implemented unicode console handling for Windows. This works
by reading/writing to the raw console handle using
``{Read,Write}ConsoleW``.
The problem is that we are going to ``dup2`` over the stdio file
descriptors when doing ``FDCapture`` and this will ``CloseHandle`` the
handles used by Python to write to the console. Though there is still some
weirdness and the console handle seems to only be closed randomly and not
on the first call to ``CloseHandle``, or maybe it gets reopened with the
same handle value when we suspend capturing.
The workaround in this case will reopen stdio with a different fd which
also means a different handle by replicating the logic in
"Py_lifecycle.c:initstdio/create_stdio".
See https://github.com/pytest-dev/py/issues/103
This works by adding an argparse Action that will raise an exception in
order to skip the rest of the argument parsing. This prevents argparse
from quitting due to missing required arguments, similar to the way that
the builtin argparse --help option is implemented by raising SystemExit.
Fixes: #1999
As discussed in the mailing list, unfortunately this might break APIs
due to the subtle differences between new and old-style classes (see #2398).
This reverts commit d4afa1554b from PR #2179.
After updating to twisted 17.1.0, again the trial tests started to fail; didn't investigate too deep, decided to just
no longer delete "zope" modules when cleaning up after pytester because it seems more zope modules keep
global state that shouldn't be discarded
When using both --last-failed/--lf and --failed-first/--ff pytest would
run all tests with failed tests first (as if --lf was not provied). This
patch changes it so that when using both flags, only the last failed
tests are run. This makes it easier to set --ff as the default behavior
via the config file and then selectively use --lf to only run the last
failed tests.
In a recent refactoring we enabled all __future__ features in pytest
modules, but that has the unwanted side effect of propagating those
features to compile()'d modules inside assertion rewriting, unless
we pass dont_inherit=False to compile().
- pytester was creating a 'pexpect' directory to serve as temporary dir, but due to the fact that
xdist adds the current directory to sys.path, that directory was being considered as candidate
for import as a package. The directory is empty and a warning was being raised about
it missing __init__ file, which is now turned into an error by our filterwarnings config
in pytest.ini.
- Decided to play it safe and ignore any warnings during `pytest.importorskip`.
- pytest-xdist and execnet raise two warnings which should be fixed upstream:
pytest-dev/pytest-xdist/issues/133
this allows a clear addition of parameterization parameters that carry along marks
instead of nesting multiple mark objects and destroying the possibility of creating
function valued parameters,
it just folders everything together into one object carrfying parameters, and the marks.
Change XML file structure in the manner that failures in call and errors
in teardown in one test will appear under separate testcase elements in
the XML report.
In the xml report we now have two occurences for the system-out tag if
the testcase writes to stdout both on call and teardown and fails in
teardown.
This behaviour is against the xsd.
This patch makes sure that the system-out section exists only
once per testcase.
This has the benefical side-effect of not calling the original
warnings.showwarnings function, which in its original form
only writes the formatted warning to sys.stdout.
Calling the original warnings.showwarnings has the effect that nested WarningsRecorder all catch the warnings:
with WarningsRecorder() as rec1:
with WarningsRecorder() as rec2:
warnings.warn(UserWarning, 'some warning')
(both rec1 and rec2 sees the warning)
When running tests with `testdir`, the main pytest session would then see the warnings created by
the internal code being tested (if any), and the main pytest session would end up with warnings as well.